The linked pre-print[1] is very interesting - this is my first time seeing NLP content analysis and Middle Eastern studies (two fields I'm passionate about) com

The linked pre-print[1] is very interesting - this is my first time seeing NLP content analysis and Middle Eastern studies (two fields I'm passionate about) come together like this. The author does two analyses in the preprint: 1: Looking at passive vs. active voice (the author cites another paper that contends that passive voice is often used to obscure the perpetrators when describing unpalatable actions), they find "a disproportionate use of the passive voice to refer to negative or violent action perpetrated towards Palestinians." So, a pro-Israel bias. 2: They find "use of more negative and violent rhetoric in reference to Palestinians compared to Israelis." Again, a pro-Israel bias. I think it's important to be mindful of how much math-washing[2], willful or not, is taking place in this analysis. The author mentions a few steps taken to reduce the code's bias (such as blindly tagging the keyword corpus), but even the types of analysis chosen may introduce bias. I'd love to see a few groups with different affiliations attempt similar papers and see if the same results persist. I have trouble poking any holes in the analysis for #1 - all I can think of is challenging the assumption that passive voice reduces perceived culpability. I think the analysis for #2 could introduce some bias, as any definition of "negative and violent rhetoric" will be subjective, but I haven't looked at the word bank or anything yet. [1] http://web.mit.edu/hjackson/www/The_NYT_Distorts_the_Palestinian_Struggle.pdf http://web.mit.edu/hjackson/www/The_NYT_Distorts_the_Palesti... [2] https://www.mathwashing.com/ https://www.mathwashing.com/